Free-space optical interconnects for multiprocessing computers.
Item
-
Title
-
Free-space optical interconnects for multiprocessing computers.
-
Identifier
-
AAI9405533
-
identifier
-
9405533
-
Creator
-
Ha, Berlin.
-
Contributor
-
Mentors: Yao Li | George Eichmann | Herbert G. Kayser
-
Date
-
1993
-
Language
-
English
-
Publisher
-
City University of New York.
-
Subject
-
Engineering, Electronics and Electrical | Physics, Optics | Computer Science
-
Abstract
-
Various new free-space optical interconnects are studied in this thesis. The focus of the research is to design and to develop fast, reliable and efficient free-space optical interconnects for linking multiprocessing digital computers.;New approaches to realize optical crossover and perfect shuffle interconnects are presented in the first two chapters. To implement the optical crossover interconnects, two methods, one based on reflective components, such as mirrors and beam-splitters, and the other based on diffractive components, such as Fresnel zone plates, are used to provide controllable data permutations among digital processing elements. A new optical perfect shuffle, which can form a cascade network of shuffles, is also proposed. This new cascadable optical shuffle makes it possible to realize optical interconnects for multiple-instruction-and-multiple-data processing applications.;Optical implementation of binary symmetric logic functions is also studied in this thesis. The proposed optical binary symmetric logic networks provide a new means of using a set of spatially interconnected contact switches as computation tools to execute various binary arithmetic operations, such as arithmetic addition and multiplication.;In this thesis, processing element distribution methods for different interconnect topologies are also studied. A new concept to distribute processing elements in a ring array rather than in a rectangular array is proposed. Employing a pair of Dove prisms, the designed optical ring array imaging system can provide both bi-directional and unidirectional links for digital processing nodes. A reflective prism is also specially designed to perform a pair of conjugated image rotation operations for the proposed reflective ring array interconnect. Both the transmissive and reflective ring array networks offer a high speed clock-skew-free data communications among single-instruction-and-multiple-data processors.;Finally, a detailed in-depth investigation of interconnect capacity limitations due to the system aberration, diffraction, and power consumption efficiency is carried out. The analytical results are used to guide the corresponding optical system design. A particular design exercise with its aim to correct all the system aberrations is presented. To optimize the design, a computer aided lens design software is also used.
-
Type
-
dissertation
-
Source
-
PQT Legacy CUNY.xlsx
-
degree
-
Ph.D.
